Vol. 184: 231-244,1999 MARINE ECOLOGY PROGRESS SERIES
Mar Ecol Prog Ser
l
Published July 28Individual variability in growth rate and the timing of metamorphosis in yellowtail flounder
PIeurunecies ferrugineus
Hugues P. ~ e n o i t ' , Pierre
p e p i n 2 1 *'Department of Biology, Ocean Sciences Center, Memorial University of Newfoundland, St. John's, Newfoundland A l C 5S7, Canada
'Science Branch, Northwest Atlantic Fisheries Center, Department of Fisheries and Oceans, PO Box 5667, St. John's, Newfoundland A1C 5x1. Canada
ABSTRACT: We studied individual variability in growth and development rates of yellowtail flounder Pleuronectes ferrugineus larvae, and its impact on age and size at metamorphosis. Larvae were reared from hatch to metamorphosis in 3 separate temperature treatments. Temperature influenced the loca- tion and distribution of individual transition ages and lengths of larvae reared in separate aquaria, mainly due to its role in determining the mean and range of individual growth rates. Individual meta- morphic ages were negatively correlated with the average growth rate of larvae. Individual metamor- phic lengths generally increased as mean individual growth rates increased, but higher mean growth also resulted in a wider diversity of lengths. In one of the treatments, the otoliths of the larvae were stained 3 times during the larval period using alizarin complexone, allowing us to reconstruct the growth history of individuals once they had metamorphosed. The body length that larvae had achieved by 2 wk after hatch correlated negatively with the eventual age at metamorphosis. This relationship strengthened the nearer larvae were to metamorphosis. These results were attributed to serial auto- correlations in body length, and to a lesser extent growth rate, for individuals during the larval period.
Overall these results suggest that events occurring early during ontogeny that affect the size and growth rates of larvae can impact life history transitions occurring several weeks or months later. As a result, variance in the timing of metamorphosis, which has been suggested as an important determi- nant of recruitment variability, may become established soon after hatching, and individual probabili- ties of survival to the juvenile stage may also be determined early on.
KEY WORDS: Timing of metamorphosis
.
Individual variability . Growth rate.
Development rate . Otoliths . Temperature . Pleuronectes ferrugineusINTRODUCTION
During the early llfe history of the majority of fish, body size increases by orders of magnitude and adult characteristics are acquired within weeks or months.
Meanwhile, eggs and larvae suffer high mortality rates (Dahlberg 1979, Sissenwine 1984). Recruitment begins to be roughly established at the late larval or early juvenile stages (Bradford 1992), when the biomass of
'Addressee for correspondence.
E-mail: pepin@athena.nwafc.nf.ca
the cohort ceases to decrease as a result of mortality, and begins to increase due to the growth of individuals (Beyer 1989, Houde 1997).
Which factors control recruitment in marine fishes is unclear (for reviews see Anderson 1988, Bailey & Houde 1989, Cushing 1990, Leggett & DeBlois 19941, however growth is important in determining overall survival.
Differences in development and growth rates will affect the age-at-size, size-at-age and timing of early life his- tory transitions (hatch, first feeding, yolk absorption, and metamorphosis) (Chambers et al. 1988, Bertram et al. 1997), and consequently may determine which indi- viduals survive (Litvak & Leggett 1992, Pepin et al.
0 Inter-Research 1999
Resale of full article not permitted
Mar Ecol Prog Ser 184. 231-244, 1999
1992, Bell et al. 1995). Physiological (capacity for meta- variance, we also reconstruct the growth history of indi- bolism and catabolism) and environmental (tempera- viduals, using the otoliths of metamorphosed fish. This ture and food availability) factors affect transition size allows us to determine when individual metamorphic and a g e (Pepin 1991, Chambers & Leggett 1992, 1996, traits (ages and sizes) become predictable by the size Benoit 1999, Benoit & Pepin 1999). Subtle differences that larvae have achieved at a n earlier date.
in the timing of transitions, which often include large changes in behaviour, morphology, habitat and niche,
can potentially create large changes in recruitm.ent, MATERIALS AND METHODS with metamorphosis likely being the most important
(Houde 1987, 1989b, Beyer 1989, Pepin & Myers 1991, Laboratory growth rate of larvae was manipulated
Bell et al. 1995). by altering the temperature at which they were reared.
Among species and populations, metamorphosis ap- We chose this approach rather than manipulating food pears to be centered around a relatively fixed length, availability in order to minimize the problems associ- and transition age vanes as a consequence of growth ated with food deprivation, which can result in compe- variation (Houde 1989a, Chambers & Leggett 1992, tition among individuals and affect the physiological Benoit 1999). Among species, population-level stan- condition of larvae. Because the objective was to focus dard deviation in both age and length at metamor- on how variation in growth rate during the larval phosis increases exponentially with increases in their period affects the timing of metamorphosis, eggs were respective means (Benoit 1999). This suggests that reared to hatching under identical conditions (8.4"C) the variance-generating mechanism is similar for all and acclimatized to larval rearing temperatures after marine fishes regardless of their environment. A nearly h.atching.
identical power law (log-log mean-vanance relation- Six, 225 1 temperature-controlled water baths were ship) for metamorphic a g e and length further suggests used to create 3 treatments (2 baths per treatment) at that the two are not independent and that a 'target' nominal temperatures of 7" (7.3 -c O.l°C, mean
+
SD), transition length may not be a n appropriate model for 11" (11.1+
0.ZoC) and 13°C (13.2+
0.2OC) in which the timing of metamorphosis a t the individual level. larvae were reared. Three or 4 (20 1, 20 X 40 X 25 cm) This is further supported by the observation that the aquaria were placed in each bath, for a total of 7 per strength of the metamorphic age-length relationship temperature treatment. The outside walls of the aquaria among individuals is directly proportional to the mean were covered wlth black plastic to enhance the con- growth rate (Benoft 1999). trast of prey items for the larvae. Aquaria were filled The role of the environment and its interaction with with 18 1 of UV sterilized filtered seatvater, and rearing genetic a n d with non-genetic parental contributions conditions were kept static except for partial water (Chambers & Leggett 1989, 1992, Benoit & Pepin 1999) changes (3 to 4 1) every 3 d during the egg stage and in determining the traits of individuals is generally every 2 d during the larval stage. Dead eggs or larvae poorly understood (e.g. Policansky 1982). This remains a n d detritus were siphoned from the bottom of the a problem because the majority of studies that have tanks during water changes. A 17 h light, 7 h dark reared fish to metamorphosis under a variety of grow- photoperiod, representative of the natural photo- ing conditions fail to report the effects on the disper- period, was maintained using florescent lights hung sion an.d distribution of traits within populations (e.g. approximately 1 m above the tanks. Temperatures Lauren.ce 1975, Fonds 1979, Crawford 1984, Seikal were measured 3 times daily (at approximately 09:00, et al. 1986). Understanding of the timing of metamor- 13:00, and 17:OO h) using a d i g ~ t a l thermometer, with phosis at the individual level is essential because any O.l°C accuracy. Temperatures varied little and the time selective mechanisnis affecting survival and recruit- series of temperature were deemed stationary (no long ment will ultimately operate a t that level (Sharp 1987, term trends or cycles) by visual inspection.Crowder et al. 1992, Rice et al. 1993). Yellowtail flounder eggs were obtained by stripping In this study we quantify variation in age and size at 6 females from a broodstock held for 2 to 3 yr at the metamorphosis in cohorts of yellowtail flounder Pleuro- Ocean Sciences Center, Memorial University of New- nectes ferrugineus, and how this variation changes foundland (Canada). An equal contribution of eggs with the growing conditions. This direction follows from each female was used to create a pool of mixed from the idea that variance among individuals in the eggs. An aliquot of 4.0 m1 of eggs (yielding -9140 i
timing of metamorphosis is generated in a predictable 340 eggs) was placed into each of 21 plastic vials and continuous manner, which is likely a result of (20 ml) that were placed in an 8.4"C bath for 20 min.
individual variation in growth rates (Benoit 1999). To After acclimation, fertilization was performed by mix- explore the idea that individual differences in larval ing 0.2 rnl of mixed sperm (from 5 males) with the eggs.
growth trajectories are responsible for generating t h s A mixture of gametes from several parents was used to
Benoit & Pepln Growth rate a n d metamorphosis in flounder 233
randomize possible parental effects within and among treatments. Filtered seawater (5 ml) was added to the vials in order to activate the sperm. After 2 min the fertilized eggs were gently poured into each of the 21 aquaria containing 18 1 of UV-sterilized filtered sea- water.
Fertilized eggs were kept at 8.4 2 0.2"C throughout the incubation period. At the first water change, 100 m g 1-' of streptomycin and 60 mq I-' of penicillin were added to the tanks Eggs began hatching 6 d after fertilization, with median hatch occurring on Day 7.
On Day 9 after fertilization, 25 to 40 larvae from each aquaria were removed a n d measured for standard length under 250x magnification using a n imaging sys- tem (Bioscan OPTIMAS@ 4.10) in order to establish the initial size distribution of larvae. To minimize distur- bances, eggs and subsequently larvae were reared in the same aquaria. Beginning on Day 9, temperatures were slowly changed (1°C d-l) to establish the 3 separate temperature treatments (7, 11, a n d 13°C).
In addition, 0.5 1 of golden algae Isochrysus sp. along with very light aeration were added to each aquarium.
Aeration was subsequently increased 2 wk later when a second antibiotic treatment was given to the larvae.
Larvae were fed ad libitunl throughout the experi- ment. Cultured rotifers Brachionus sp. enriched with a mixture of powdered media a n d fresh algae were used throughout the experiment, although the size of rotifers added to the tanks depended on the size of the larvae. The average size of rotifers added to the aquaria was increased over the larval period by screening the rotifers using a nylon sieve. Rotifers mea- suring 40 to 120 pm were added twice daily at densities of 4.0 ml-' beginning 1 wk after hatching. Once aver- a g e larval size reached 4 to 5 mm, the mean rotifer size was increased by using only rotifers > l 0 0 \pm. For lar- vae between 5 to 6 mm, rotifers were selected using a 150 pm mesh. In addition to rotifers, brine shrimp Artemia sp. were added in densities of 2.0 ml-' once larvae reached an average of 7 to 8 min. Densities of brine shrimp were increased to 4.0 ml-', a n d rotifer den- sities decreased to 2.0 ml-' approxin~ately 1 wk later.
As larvae approached metamorphosis, individuals were examined every second day in order to find those that had metanlorphosed. We defined metamorphosis as the point at which the iris of the migrating eye reached the neural crest and was visible from the other side of the body. This is commonly used as a standard method of scoring nletamorphosis (e.g. Chambers &
Leggett 1987, Bertram et al. 1997). Metan~orphosed individuals were removed from the aquaria, eutha- nized a n d preserved in 95% ethanol. To assess the effect of preservation on body measurements, stan- dard length and maximum body depth (measured perpendicular to the longest axis of the fish) were
measured on >350 live individuals under 64x magnifi- cation, prior to preservation. These individuals were remeasured 3 to 4 mo later ( w h e n all metamorphosed fish were measured). Age at metamorphosis is defined as the number of days between scoring a fish a s a metamorph a n d modal hatching date.
Experiments were terminated prior to metamorpho- sis for 4 fish at 7°C. The larvae were nonetheless mea- sured a n d their development was r n i i r ~ h l y s r n r ~ d h y the position of their migrating eye. These individuals were included in some of the analyses that a r e able to accommodate such data (see 'Analyses' section for details).
Individual growth histories. Given the inability to tag individual larvae, a n d the difficulty in rearing indi- viduals in isolation to make repeated length measure- ments (see Bertram et al. 1997), we chose to utilize the relationship between body length a n d otolith diameter of common a g e fish to reconstruct the growth history of individual fish. In order to alleviate problems associ- ated with non-daily deposition of otolith rings ( e . g . Sogard 1991, Szedlmayer & Able 1992) a n d variation in the oto1ith:somatic growth relationship (Campana 1990), we marked larval otoliths a t specific time intervals in the 11°C treatment. We could then confidently mea- sure the diameter of the otolith at that time and back- calculate the size of the fish. We used a solution of sea- water and alizailn complexone (Cl9Hl5NO8 H 2 0 ) , which is incorporated into the n e w otolith growth increments a n d fluoresces under UV light. Otoliths were marked on the 18th, 30th and 43rd day of the larvae period in 3 of the 7 tanks in the 11°C treatment (Table 1).
To mark the larval otoliths, the water volume in the 3 aquaria was reduced to 5 1. Four liters of a 112.5 nlg 1-' alizarin solution (prepared by diluting 1.35 g of alizarin in 270 m1 of distilled water a n d topping up with 11.7 1 of seawater containing algae a n d rotifers) was gently poured into each aquarium to yield a final concentra- tion of 50 nlg 1 - l . Immersion of larvae lasted 16 h, after which they were transferred using a nylon filter to aquaria containing filtered seawater a n d algae. Over- all the staining process a n d subsequent transfer of larvae to clean water resulted in very low mortality ( < 3 individuals per tank per staining period).
These aquaria were checked for metamorphosing larvae in the same manner as unmarked treatments, a n d individuals were preserved similarly. After mea- suring for preserved length, left a n d right otoliths (sagitta, lapilli, and asteriscii) were removed using a dissecting scope a n d mounted on microscope slides using thern~oplastic cement. Lapilli were viewed under 1000x magnification using 365 n m UV light a n d a wide band interference blue filter. The diameter of each flu- orescent ring a n d the total diameter of the otolith were measured along the longest axis using an ocular micro-
234 Mar Ecol Prog Ser 184: 231-244, 1999
Table 1. Pleuronectes ferrugineus. Number of larvae (n) surviving to increase in otolith diameter (pm) divided by metamorphosis and larvae (c) that were removed from the experiment the time interval over which arowth was prior to metamorphosis in each tank at each temperature, along with the measured, Growth rate from hat& to Day 18 (*SD) mean ages and lengths at metamorphosis, and ( r ) Kendall's tau
nonparametric correlation between a a e and lenqth. All valut!s of .r were assuming a 'Onstant diam.eter statistically significant ( p < 0.05) except where indicated (ns). Note th.at of 11 pm at hatch (estimated from the diame- no larvae survived in Tanks 6 and 7 at 7"C, and larvae were only avail- ter:length relationship for the youngest lar- able from 1 tank in the 13°C treatment. '. tanks at l l ° C in which otoliths vae, and assuming a mean hatch length of
were stained
3.2 mm). Analyses involving otolith diame-
Temp. Tank n c Mean age SD Mean length SD Kendall.'~
("C) (days) (mm) 't
7 1 25 2 118.0 14.5 17.1 2.3 -0.41
2 84 118.0 10.7 17.2 2.0 0.24
3 17 1 121.0 10.6 15.8 1.7 -0.09 ns 4 13 121.8 16.0 15.3 2.4 0 . 5 8 5 4 1 110.0 8.7 16.2 1.5 -0.91 ns
11 1 195 111.9 16.6 16.9 2.3 0.52
2 ' 62 81.2 9.5 18.0 2.2 0.23
3 161 97 5 15.7 16.4 2.0 0.32
4 ' 104 8 6 6 12.0 1 8 0 2.1 0.21 5 36 86.0 15.0 19.0 2.8 -0.28 ns
6 167 93.8 13.8 17.0 2.3 0.41
7' 41 75.1 10.2 19.7 3.2 0.16ns
13 1 71 75.4 11.9 18.9 2.2 -0.05 ns
ters and growth rates were performed using rank correlations (which are independent of absolute magnitude).
Analyses. Covariance between age and length at metamorphosis among individuals was analyzed using nonparametric correla- tlon analysis (Kendall's
z),
which assumes a monotonic relationship between variables, in addition to segmented regression analysis that fits a biphasic regression to the data. A nonlinear, segmented fitting algorithm (Wil- kinson 1992) was used to test the working assumption that the data can be represented by 2 linear segments, differing in slope. In particular, we were interested in determining non-arbitrarily a point at which age and length at metamorphosis go from being un- meter. Fluorescent bands did not appear on the aster- correlated to being positively related. The model used iscii, likely because these form later during ontogeny was of the form:(Secor et al. 1992). Sagittae had large accessory length = B. + B,(age), when a g e
<
BREAK primordia that appeared to form once metamorphosislength = B.
+
(B1+
B2)(age), when age > BREAK (1) began (H.P.B. unpubl.), and that would have requiredsubstantial polishing in order to measure the bands.
In order to establish the relationship between otolith diameter and larval length for each staining period, 10 to 15 larvae were removed from each of the 'stained' tanks the day following each marking and were measured and preserved individually in ethanol.
The otoliths from these larvae (lapilli and sagittae) were later removed, mounted and measured as described above. Unfortunately the larvae that were taken for the 18 a n d 30 d staining periods were pre- served in ethanol that was too weak (<80%) and consequently became too acidic to properly preserve the otoliths. We later o b t ~ i n e d 11 and 20 d old larvae from 2 separate general. rearing stocks held at the Ocean Sciences Center (Memorial Un~versity of Netv- foundland) that had similar mean lengths as the larvae from the experiment a t 18 a n d 30 d respectively in order to establish otolith diameter:body length rela- tlonships. There is the potential that different growth rates of these larvae would result in differences in the slope and intercept of the relationship (Reznick et al.
1989, Campana 1990). In order to avoid introducing any biases in our data, all analyses were performed using otolith diameters rather than back-calculated body lengths. Growth rates were calculated a s the
where B,, is the intercept (mm), B1 is the slope of the first segment (mm d-l), B2 is the difference in slope between the first and second segment (i.e. the slope of the second segment = B,
+
B2), and BREAK (days) is the inflection between the first and second segments.Confidence intervals (95 %) are reported for each para- meter. We acknowledge that the relationships to which we fit the lines are continuous, but as mentioned pre- viously, we were most interested In determining the location of a possible break point, rather than testing hypotheses on the magnitudes of the slopes or describ- ing the functional relationship in detail. We only fit biphasic regressions to data from those tanks that visu- ally showed a change in age-length correlation over time. Robustness of parameter estimates was tested by varying the initial values used in the estimation.
The analyses of the frequency distributions of both age and length at metamorphosis were performed using event analysis (Cox & Oakes 1984, Chambers &
Leggett 1989) in which the entire distnbution of ages or lengths in a population is used to describe the timing of metamorphosis. Event analysis fits a theoretical statistical distribution to the observed frequency data, and uses the parameters that describe that probability distribut~on (its shape and position) to test for differ-
Benoit & Pepin: Growth rate a n d metamorphosis in flounder 235
ences among different observed frequency distribu- tion. Using this technique, one simultaneously tests for differences in the mean, dispersion a n d distribution of metamorphic a g e s and lengths among treatments (i.e.
compares all of the individuals in the population, not just the average ones). This is one of the advantages of this technique over standard parametric analyses, which compare average responses and use variance as a description of error, as information from all indi- viduals in the population is retained at its original level, rather than amalgamating it in the estimation of a mean. A second advantage is that this type of analy- sis allows the incorporation of data on individuals that were removed from the experiment prior to undergo- ing the event (termed 'censored' individuals). This reduces potential biases because until the individuals were removed from the experiment they were at 'risk' of undergoing the event a n d consequently they con- tribute to the dispersion. For a more detailed explana- tion, a s well a s examples of the application of event analysis to the study of fish early life history, see Chambers & Leggett (1989) a n d Pepin e t al. (1997).
A Weibull distribution, which is described by 2 para- meters, was used to model individual ages a n d lengths at metamorphosis as it provided the best log-likelihood fit to the data. The survival function of the Weibull distribution is defined as
where t is the tinling of the transition (i.e. individual a g e or length a t metamorphosis), a is a scale parameter in units of time (days) or length ( m m ) , a n d y is a di- mensionless shape parameter. The survival function is the complement of the cumulative frequency distribu- tion [F(t)] of metamorphic ages a n d lengths [i.e. S(t) = 1 - F(t)]. Eq. (2) can be modified into a n accelerated failure time model by including ' n ' treatment variables (e.g. temperature, fish density, etc), described by a vector v' = (v,, v,, ..., v,,), such that
S(t; v) = exp [-(tla)' exp"'B] ( 3 ) assunling that the elements of v interact through a linear function, v'p, where
p,
a r e parameters to be estimatedThe analysis uses log-transformed ages or lengths, which yields the standard extreme value distribution that is multiplied by a scale factor (analogous to the variance of a normal distribution), and translated by a location parameter (-mean). Observations of individ- ual event times a r e related to explanatory variables in a regression model, as
where Y is the vector of log-transformed a g e s or lengths at metamorphosis, V is the matrix of explana-
tory variables,
P
1s the vector of coefficients to b e estimated, o is the scale parameter to be estimated, a n d E is the extreme value distribution in the case of the Weibull distribution. Sigma (o) is the inverse of the shape parameter y in the Weibull distribution, a n dp , ,
a n estimate of the intercept, is the location para- meter of the extreme value distribution [i.e. = log(u)].Finally, E is a random error vector from the assunied extreme value distribution. In E q . (4) the explanatory variables act multiplicatively on individual ages/sizes, resulting in acceleration of event timing. Parameters were estimated using iterative maximum likelihood methods.
RESULTS
Overall, survival was variable across aquaria a n d temperatures (Table 1). As a result only 5 tanks w e r e available for analysis at 7"C, with generally low numbers except in 1 tank. At 11°C survival was vari- able across replicates but sample sizes in all aquaria were high. Due to a n accident in the 13°C treatment 2 mo into the study, all but 1 replicate were lost. As a result of this variability we will include all of the aquaria in a general analysis of temperature a n d growth effects, but w e will subsequently focus on the 11°C treatment to assess sources of experimental error among replicates.
Effects of preservation
Preservation in 95% ethanol caused substantial shrinkage of the metamorphosing larvae. Preservation effects were estimated by fitting an asymptotic rela- tionship to data from 371 fish, where
preserved length =
P,
( l - exp[-P, fresh length]) (5) wherep,
is the a s y n ~ p t o t e of the curve, a n dP,
deter-mines the rate at which the asymptote is reached.
This model performed very well (R2 = 0.96, p i 0.0001), with estimated parameter values a n d associated asymptotic standard errors of
P,
= 161.03+
26.14 a n dp,
= 0.0061*
0.0011 ( d u e to the high degree of explained variance w e do not present a plot of these d a t a ) . Preservation effects were also estimated for measurements of body depth(P,
= 43.47*
6.93,p,
= 0.025*
0.004, R2 = 0.89, p < 0.0001). The high degree of explained variance suggests that differential shrinkage of larvae (resulting in changes in the size rank of larvae) did not significantly affect our results regarding the statistical distributions of metamorphic lengths. All subsequent analyses have been performed using preserved lengths as the unit of study.236 Mar Ecol Prog Ser 184: 231-244, 1999
Covariation between age and length at metamorphosis
Mean age and length at metamorphosis were negatively correlated across tanks (n = 13, r = -0.85,
p < 0.0001) (Fig. 1). The relationship is independent of
temperature, suggesting that the average response of a population of larvae is a result of the mean growth rate.
Slower growth results in increased age and decreased length at metamorphosis.
Within aquaria the relationship between age and length is complex (Fig. 2 ) . In general there is little or no covariation between a g e a n d length in treatments with survivor numbers less than 85 ind. tank-' (Table 1).
Some exceptions occurred for tanks at 7°C but these appear to be driven by relatively few older larvae (>l40 d of age)(Fig. 2). In contrast, a positive covari- ance develops in tanks with higher densities (Fig. 2, Table 2). In Tanks 1, 3, 4 , a n d 6 of the 11°C treatment a positive correlation between age and length at me- tamorphosis appears after an average of 89.0 to 97.9 d , as determined by segmented regression analysis (Table 2). By that time almost all of the larvae in the lower density tanks (Tanks 2, 5 and 7) had undergone metamorphosis, and consequently do not show this tendency.
Age at metamorphosis
It is difficult to describe the distributions for the raw data on metamorphic age beyond the mean and vari- ance (Fig. 2). Nonetheless, Weibull distributions fit rea- sonably well to observed distributions. Event analysis reveals that age at metamorphosis differed signifi- cantly among temperature treatments (Table 3), and a
Mean length at metamorphosis (mm)
Fig 1. Pleuronectes ferruqineus Relationship between mean
a g e and length at metamorphosis in ind~vidual aquarla
(temperature and replicate) ( 0 ) 7"C, ( A ) 11°C and (0) 13°C
Age at metamorphosis (days)
Fig, 2. Pleuronectes ferruqineus. Ages and lengths at meta-
morphosis for each a q u a n a of laboratory-reared yellowtail
flounder Each point In the bivariate plots represents the
response of 1 individual (see Table l for sample sizes). Fre- quency distributions a r e for the univariate responses of a g e and length. T h e scales for the distributions can b e found on
the secondary X - and )/-axes for age and length respectively.
Note that Tanks 4 and 5 from the 7°C treatment are c o m b ~ n e d In a slngle panel, with additive frequency bars
Benoit & Pepin Growth rate a n d metamorphosis in flounder 237
-P
Table 2 Pleuronectes ferruqineus. Results of segmented regression analyses on individual measures of a g e a n d length a t meta- morphosis for those tanks that display a biphasic relationship. Parameter estimates are given with lower a n d upper confidence
intervals. See 'Materials a n d methods' for details regarding the analyses
Temp. ("C) Tank df R2 (corrected) Intercept Slope 1 Slope 2 Break
P
11 1 4, 191 0 . 5 3 22.4 (17.3, 27.5) - 0 08 (-0 1 4 , - 0 03) 0.24 (0.17, 0 30) 97 9 (93.6, 102.1) 3 4, 157 0.41 19.2 (16.3, 22.2) - 0 0 5 (-0 0 8 , -0.01) 0.21 (0.16, 0 27) 9 5 . 9 (91.9, 99 9 ) 4 4 , 100 0.29 18.6 (14.7, 22.5) - 0 02 (-0 06. 0.03) 0.22 (0.12, 0 3 1 ) 89.0 (87 9 , 90.1) 6 4 , 163 0.45 1 3 6 (10.1, 17.1) 0 . 0 3 (-0.01, 0.07) 0.17 (0.09, 0 2 4 ) 96.2 (90.3, 102.11
Table 3 Pleuronectes ferruqineus. Fit to the accelerated fallure model (Eq. 4 ) for log-transformed metamorphic ages a n d lengths
Temp Age a t metamorphosis Length a t metamorphosis
("C) df Estimate SE
x2
P Estimate SEx2
PIntercept 1 4.27 0.02 57039.9 <0.0001 3.04 0.02 28619.2 <O 0001
Temp 7 1 0.460 0.021 498.0 <0.0001 -0.116 0 . 0 2 1 32.0 <0.0001
1 1 1 0.101 0.026 29.6 < 0 . 0 0 0 1 -0.013 0 . 0 1 9 0.50 0.48
1 3 1 0 0 - - 0 0 -
Density 1 1.78 X I O - ~ 0.90 X I O - ~ 416.0 <0.0001 -0 8 3 X I O - ~ 0 9 0 X 10-S 8 6 . 2 <0.0001
Scale 1 0.140 0.003 0.141 0 0 0 3
Log-likelihood: 441.1 503.5
significant effect of fish density was also detected. Age at metamorphosis increased as temperature decreased and as the density of fish at metamorphosis increased.
Variance in a g e at metamorphosis was not signi- ficantly related to average a g e , as determined by a power law, or log-log linear regression = 2.13, p > 0.1, R2 = 0.15), contrary to the findings of Benoit (1999) for other species. As noted previously, event analysis had suggested that larval density affected a g e dispersion, consequently its effect was added to the linear model relating mean and variance. Although this increased the explained variance to 4 8 % , all para- meters were not statistically significant at or = 0.05.
Density did however significantly affect the skewness of the a g e distribution for untransformed data (r = -0.73, p < 0.007, n = 12), with lower density popula- tions showing positive skewness and high densities resulting in slight negative skewness (Fig. 3 ) .
Length at metamorphosis
Event analysis shows that length at metamorphosis differed significantly among the 7°C and the 11 and 13°C treatments, but the latter two did not differ signif- icantly from one another (Table 3 ) . A significant effect of larval density was also detected. Overall, length at metamorphosis decreased with decreasing tempera- ture and with increasing density. Given the strong density effect at l l ° C , we explored the impact of larval density on the length distributions within each tem- perature treatment. Furthermore w e incorporated the
covariance with a g e to explore its effect on shaping the distributions of length within each treatment, as the segmented regression analysis had shown that this effect was not constant across aquaria and tempera- tures. The intercept, or location of the distributions, increased as temperature increased (Table 4 ) . Age did not significantly affect the distribution of sizes at 7"C, but had a significant impact in the other 2 treatments, although the effect at 13°C is largely d u e to a few indi- viduals (Fig. 2). The effect at 11°C reflects the unequal distribution of ages among lengths (largest individuals also oldest) at higher densities. Density effects were marginally insignificant at 7°C but highly significant at 11°C (Table 4 ) . Density effects at 11°C were largely
Density
Fig. 3. Pleuronectes ferrugineus. Effect of fish density on the skewness of the metamorphic a g e distribution in each tank.
( 0 ) ? " C , (a) 11°C a n d (0) 13°C
Mar Ecol Prog Ser 184. 231-244, 1999
Table 4. Pleuronectes ferrugineus Temperature-specific fit vidual characters, without greatly affecting the range of the accelerated failure model (Eq 4) to log-transformed and dispersion of the lengths in a population,
metamorphic lengths, including the effects of age and density
Parameter SE Chi-sq P estimate
7°C
Intercept 2.67 0.11 602 <0.0001 Age (x10-7 1.50 0.98 2.36 >0.1 Density ( x ~ O - ~ ) 0.61 0.327 3.48 0.062
Scale 0.11 0.007
11°C
Intercept 2.74 0.02 18071 <0.0001 Age (x10-') 4.55 0.23 386 <0.0001 Density ( X I O - ~ ) -1.97 0.09 457 <0.0001
Scale 0.12 0.003
13°C
Intercept 2.75 0.085 1036 <0.0001 Age (x10-~) 3.26 1.13 8.29 <0.005
Scale 0.11 0.010
manifested in a shifting of the mode of the distribution towards smaller sizes a s density increased, as the absolute range of sizes was relatively constant across aquaria (Fig. 2). The effect of density on the mode of the distribution contributed to a significant positive relationship between density and skewness in length (r = 0.84, p < 0.0001, n = 13) when all tankdtempera- tures are considered (Fig. 4).
A statistically significant power law, relating log- transformed mean length and its associated variance, explained 52.7% of the variance in the dispersion of lengths (slope = 3.66
*
1.00 (SE), intercept = -3.85*
1.24,F,,,, = 13.38, p < 0.005). Larval density had no signifi- cant effect on dispersion = 1.90, p = 0.20). This result further supports the observation that density affects mainly the Skewness of the distribution of indi-
Density
Fig. 4. Pleuronectes ferrugineus. Effect of fish density on the skewness of the metamorphic length distnbution. in each
tank. (0) ir°C, ( A ) 1 1°C and ( 0 ) 13°C
Body depth
Length of metamorphosing larvae explained 87.6 % of the variance in body depth in a log-log regression (F1,1029 = 2578, p < 0.0001). Expanding the linear model, we found a significant tank (nested in tempera- ture) effect (F16,1003 = 15.22, p < 0.0001), although both temperature and length X temperature interaction effects were not significant = 1.16, p > 0.3, and F3,1003 = 1.93, p > 0.1, respectively). Despite the signifi- cant tank effect, the larger model increased the explained variance by only 3.6%. This result means that overall, body depth (a proxy for the mass of indi- viduals) at metamorphosis did not significantly vary independently of metamorphic length. Furthermore, the lack of significant interactions between body depth and temperature suggests that mass scales to length in the same manner under different growth regimes (i.e.
individuals are not growing in length at the expense of body mass).
Otolith reconstruction
Length corresponds reliably to otolith diameter in common aged fish as small as about 5 mm (Fig 5 ) . Although the larvae from the 2 youngest groups were grown under different conditions than the experimen- tal fish, they nonetheless fit in with the overall trend of increasing otolith diameter with increasing body size.
This allows us to use otolith diameters for 18 and 30 d old experimental larvae as indices of larval length (i.e.
we can assume that a relatively larger otolith came from a larger larvae, but w e cannot estimate a length precisely). Overall the correspondence between otolith diameter and body length improves as mean length and the range of lengths increases, as suggested by decreasing standard errors and increasing explained variance (Fig. 5, Table 5). The relationship at metamor- phosis is weaker (Table 5 ) , likely due to the diversity of ages-at-length, which can affect the otolith diame- ter:body length relationship (Rezn~ck et al. 1989, Cam- pana 1990).
The relationship between measured diameters for left and right otoliths of experimental fish improved as the larvae aged and grew (Table 5). For 18 d old lar- vae, the diameter of the left otolith explained 64% of the variance in that of the right, which is likely a result of measurement error for small otoliths, in addition to the lim~ted range of di.ameters being compared.
Explained variance increased as dld the slope of the
(Fo'O) t6 '0 (6
0'
~) ICE 16'0
Z'OCZ - 9'POT Z8'9P1
~1 1 s~soqd~owelaur
IV
(~
0' 0) 86'0 (~
6' 0) 6C.I t6'0
0'68
0'9'2 -
S8'PS L11
(pau!els) p10 P
&P
(170.0) 68'0 (6E't)
EZ'P P8'0
S'SP - 0'82 90'tE
tl I (pau!els] p10
oc
P P 81 (pau!elsI p10 t11 ~ .8 1c Z O'L - C S'E '0 ~9 ) 60.8 P'L (So zt'
0) 0. (~(3s)
ad ol
(3s)
~ ldasialur1 8 (urw) a6ue8
(ww) lalawerp ue
an N
(l al aw e~
p ljal
adols X
+
~d as ia lu
~ =
lalaure!p lq6u) sqlyolo ljal
pue 1q 6 !1
uos!iedwo3 JO
(10'0) OL
'O
(~
6' 0) OZ E 6S'O
P'ZE - 6'ZI Z2'81
981 s~soqdioure~aur
IV
(10.0) EI
(OP'OI 'O
z
PL~6 .0 S'P
- I z6.9
~0 .0
o
1 E (pau!els) PPP C P
(~
0' EZ'O 0) (PS'O)
P9-0- P8'0
LL.6 - 66'P PO'L
0 P ,uPIO P O
Z
(10'0) 01'0 (9z.0) 80Z
Lf'O L0.S
Z O -
' P 6P'P
82 .PI0 P
I1 -
- -
(3s)
a d o
~
(3s)
~ lda31alurZH (urw) ahuea
(urur) ql6uar u e a
N ~
(lalaue!p ylgolo
adols X
+
ldas~alu!
= 7s ) y lb ua
~ j o uo~lernsres-yseq
su IOJ
o~
ss al ba
~
s!soqdioruelaur
le pue pouad 6u1u1els
qsea l e sql~lolo lq6!i
pue ljal jo sialaure~p aql Bu~ieduros
suotssai6ai pue 's~soqdrowe~aur
le aenieI
pue aer\lel pa6e uoururo~
~ O Jq~
hu al piepuels pue ialawe!p
ql
~~
ol
o 6u
qe 1a
~ suo~ssaihax
~snau.r6nrraj sapauornald S
a[qeL
s~soq&oure)aur le a6e aleledas li[a~!lsajja
am
6u!op OS
le yl6ua1 pue a6e
uaaMlaq d1ysuoge1al aql 'alojaq
Lg '(p9 '61j) s!soqdrourelaur
le az!s aql
'd!qsuo!le[a~ 01
palou am se ' l a
~ a
~ o
~ 61 ~6 '(
) .[P
la urelllag L q
punoj
p
&P e le a z!
s snslaA a6e
ay] wolj slenprsai aqi
pale1 leql
03 JPIILUIS
ilnsai e 's!soydrourelaur le
a6e u1 asue
-ail03 aM '$Insal s !y l jo 6u!puelsiapun lallaq
u1e6 o~ P
-UPA aql jo
%
0 1 pt
lnoqe pau!eldxa 8s
lalaure!p qi~lolo
'(z
. 61 pue , 1)
' z
p sy ue~
yeaM seM UI
s!soqdloure~aur
&v
ppue 'pa6e Aq
ae hq se pa~oldru! uo!le[al~os
aqL
'(
a1qeL 9
'g p 81 euenbe [Ie IP
loj palDalap seM
yuel ysea loj s
~s oy d~
or ue
~a u~ le abe pue p le l a~
au e~
p
qlqolo .6u!lelai suo1ssai6ai woij
paleInzqe3
slam
slenp!sax
.s!soqdrourelaur l e ql6ua1 pue
laued urolj (2)
s~
en p~
sa
i aql
uaamlaq d!qsuo!~e~al (p) '&
. 11 le 'L
(0
PUP )
P (0) 'Z
(0)
sy ue
~.
loj 'abe
p ~p JO
(s) pue '01 ( '81 q)
(P
le )
ialawerp ylqolo
pue s!soqdiou~e~au
~e a6e uaamlaq
d!
ys uo
~~
el a~
-g '6 1~
s!soqd~oure~aur le a6e pue (y16ua1 Lpoq-)
lalauIe!p
ql11olo uaaMlaq uo!le[alios
a~
11
~6 au lues!j;u61s
v
'sql~loio ljaI pue 1q61~
qloq uo apeur alaM
sluauralnseaur uaqhz uasoqs L1uropuel
seM sqi11olo
ay1
JO
'~ -[
01 10
~ leyl sasA~eue
aql 109 .(elep .[qndun
'8 .d '~
) ~l pl 0U l~
d [[PLUS
UO 3 0 l~
PL Ul
aql A OJ
q paSnP3
qpdr?-[ aql uo sa~l!ur.~ojap 6u1sealx11
anp Allled 01
's!soqdrourelaur le painseaur lalauretp [elol
ay1 ioj
Al~q6!ls paseaisap asueueA p
au ye
~d x~
.aealel p1 0 p
&p
pue
1oj OF
jo anler, (1
e splemo$
6u1pual) d!ysuo!le[al
aehlel p10 P
&P
pup (v) lualajj!p woij 'sassels a612 (01 '1 I (0)
oz
aen1eI ~ o
j ql6ua1 Apoq
plepuels pue ialaure~p yl~rolo
uaamlaq d
~q su o~
le
~a x -S .61g
~-
iapunorj 6 EZ
U !
s!soqdiourelaw pue ale1 qlm013
:u!dad 9 qouag
240 Mar Ecol Prog Ser 184: 231-244, 1999
Tank n Age-L18 Age-L30 Age-L43 Age-LM Age(residua1)-LM
2 50 -0.39" -0.54 "' -0.62"' 0.16 ns 0.37 "
4 97 -0.51"' -0.6"' -0.75"' 0.2' 0.51"'
7 39 -0.58"' -0.59"' -0.74 " ' 0.22 ns 0.22 ns
7' 38 -0.62." -0.72"' -0.86"' 0.14 ns 0.16 ns
All 186 -0.41 "' -0.55"' -0.63"' 0.08 ns 0.41 "'
Length autocorrelation Growth rate autocorrelation
A0,18- A18,30- A30.43-
Tank n L18-L30 L30-L43 L43-LM A18.30 A30.43 A43,meta.
2 50 0.75"' 0.89"' 0.12 ns 0.11 ns 0.63"' 0.34.
4 97 0.88"' 0.91 "' 0.12 ns 0.28" 0.54"' 0.20'
7 39 0.75"' 0.92"' -0.03 ns 0.17 ns 0 7"' 0.4 '
7" 38 0.84"' 0.91 " ' -0.12 ns 0.21 ns 0.68"' 0.36'
All 186 0.81"' 0.91 " ' 0.19' 0.21" 0 59"' 0.16'
"Strong outlier (length = 32.4 mm) removed
Table 6. Pleuronectes ferrugineus. Spearman rank correlations (and their statis- size autocorrelations are only partly
t ~ c a l significance) for measurements made on otoliths stained with alizarin corn- due to autocorrelation In growth rate
plexone Age is the observed age at metamorphosis, length (Lx) is the d ~ a m e t e r fast growers being more likely to
of the otolith at age ' X ' or at metamorphosis (LM), and A a , b is the growth
ratefrom time 'a' to time 'b' ns: p > 0.05, '0.01 < p < 0.05, "0.001 p < 0 01, quickly a t later tilnes).
"'p < 0.001 early growth (and the size achieved)
may be important in determining how larvae rank in length relative to one another at later times.
Lastly, it should be noted that al- though aquaria in which larval otoliths were stained produced larvae that were on average slightly younger and larger at metamorphosis than other tanks at 11°C (Table l ) , this is largely a n effect of fish density (as discussed in the previous sections). Because 10 to 15 individuals were removed from these tanks during each staining period (in order to establish the SL:otolith diameter relationship), den- sities tend to be lower than the other tanks, and consequently growth rates a little higher.
into that which is due to the growth rate up to 43 d and that due to the deviation (residuals) from the meta- morphic age that would be predicted from that rela- tionship. The significant positive correlation between length at metamorphosis and 'residual' age at meta- morphosis in Tanks 2 and 4 (and for the data overall) (Table 6) suggests that larvae that undergo metamor- phosis sooner than would be predicted from their size at 4 3 d (negative residual) do so at a size that is smaller than the overall average of 18.2 mm. On the other hand, those larvae that took longer than expected to metamorphose were larger than average. When w e combine this positive relationship between 'residual age' a n d metamorphic length with the negative rela- tionship between length at 4 3 d and age at meta- morphosis, the two effectively cancel each other out, leaving metamorphic age and length uncorrelated.
The predictability of metamorphic age by the body length that individuals have reached as early as 18 d into larval life, in addition to the observation that meta- morphic age becomes increasingly predictable as lar- vae grow, su.ggests that consistency must exist in the relative size rank among individuals. Strong autocor- relations in length were detected using rank correla- tions (Table 6). Size at 43 d of age did not correlate strongly with length at metamorphosis as mentioned in the previous analysis. Rank correlations of growth rates from 0 to 18 d versus that from 18 to 30 d (r = 0.11 to 0.28) and growth over the 18 to 30 d period versus that for 30 to 43 d (r = 0.5 to 0.7) suggest that the strong
DISCUSSION
Growth rate proved to be important in determining the timing (age and size) of metamorphosis both among and within populations of yellowtail flounder, consistent with an earlier synthesis of existing data on other species (Benoit 1999). Although temperature may determine the growth potential of larvae, affect- ing the location or mean of the distribution of meta- morphic ages and sizes (Fig. l ) , the considerable dis- persion about the mean suggests that individual characters vary in a continuous manner. If this is the case, then averaging the metamorphic traits of indi- viduals in comparisons of different rearing environ- ments (e.g. Crawford 1984, Seikai et al. 1986, Minami
& Tanaka 1992) may mask individual level responses that vary continuously with individual growth trajecto- ries. Below, we present evidence that this may be true.
Because individual average growth rates (mm day-') are not independent of either age or length at meta- morphosis, any hypothesis testing and estimates of explained variance are inappropriate, however, the slopes and intercepts of the relationships are nonethe- less valid estimates (D. Schneider, Memorial Univer- sity of Newfoundland, pers, comm.). Consequently the relationships between growth and age or length at metamorphosis can be compared across environments.
In our study, increasing growth rates decreased the age at whlch individuals underwent meta.morphosis (Fig. 7 ) and the relationship was continuous across
Benoit & Pepin: Growth rate and metamorphosis in flounder 24 1 - -
temperatures and tanks, suggesting that growth rate averaged over the entire larval stage roughly deter- mines age at metamorphosis. The effect of growth rate in determining length at metamorphosis is not a s sim- ple. Overall the length at metamorphosis increased with growth rate as did the range of lengths for a given growth rate. As growth increases, the increasing range of lengths might suggest a dependence of individual lengths on the growth rate experienced by larvae later during ontogeny, once age at metamorphosis has been roughly established (as in the present otolith study).
For a given time interval prior to metamorphosis, a 10% change in growth rate will have a greater ab- solute impact on final size for larvae that were growing rapidly, resulting in a greater range of lengths.
Data from the otolith reconstructions further suggest that although average growth rates may roughly de- termine age and length at metamorphosis, precise pre- dictions of the timing require examination of the vari- ability in daily growth experienced by individuals over the course of their larval life. Although strong auto- correlations were detected in the sizes of larvae over time, this was only partly d u e to autocorrelation in growth rates. Such size autocorrelations have also
10 1 l I
0 05 0.15 0.25 0.35 0.45 Average growth rate
(rnrn'day")
Fig. 7. Relationship between the growth rate of individual larvae, averaged over the entire larval period, and (A) age, or
(B) length at metamorphosis. (0) 7, ( - ) 1 1 and ( A ) 13°C
been found in larval red sea bream (Umino et al. 1996), juvenile turbot Scophthalrnus maximus (Rosenberg
& Haugen 1982, Imsland et al. 1996) and halibut Hippoglossus hippoglossus (Hallaraker et al. 1995).
Likewise, Chambers & Miller (1995) found similar degrees of autocorrelation in larval Atlantic menhaden Brevoortia tyrannus for similar time spans between measurements (10 to 15 d ) and also found that growth autocorrelations were weaker than those for size. The weaker rank correlations for growth rate suggest that the growth rate and the absolute size achieved early during larval life may be important in determining the timing of metamorphosis. Because the first half of larval life (particularly once feeding is initiated) may be a period in which body length increases almost exponentially (e.g. Bertram e t al. 1997), this creates the potential for the establishment of difference among individuals that may persist well into the larval period.
In addition to the importance of growth rates during early ontogeny, data from the otolith diameters also suggest that the precise timing of metamorphosis depends on the growth experienced in the latter half of larval life. Individuals that underwent metamorphosis sooner than would be predicted by their growth rate up to Day 4 3 did so at a smaller size (the opposite being true for individuals that took longer than predicted).
Unfortunately we do not have an estimate of growth rate from 43 d to metamorphosis that is independent of metamorphic length, precluding us from finding what determines whether a n individual will metamorphose sooner/smaller or later/larger than predicted.
In addition to the results from the otolith study, the density-dependent effects observed in some of the tanks a t l l ° C may support the argument that growth later in larval life affects the timing of metamorphosis.
A potential mechanism for the development of a posi- tive correlation between age and length is that as den- sities in the aquaria decreased due to the removal of metamorphosed individuals, the remaining individuals experienced a decrease in competition as food was partitioned among fewer larvae (i.e. increased growth rate). The results of the otolith study suggest that these remaining individuals are the slower growing mem- bers of the population (developmentally further from metamorphosis), and thus have a longer time to utilize these benefits and grow to larger sizes. This may ex- plain why some of the largest fish at 11°C were from the higher density tanks.
Although the density-dependent effects observed a t 11°C were the result of survival variability, and not intentionally included in the experiment, their occur- rence has allowed us to explicitly quantify any labora- tory artifacts present in our study. Although such strong density effects are unlikely in more natural situ- ations (e.g. Fortier & Harris 1989), by quantifying the
